To the Editor:
The recent paper of Duan et al.(1) represents an impressive amount of data and contains many thoughtful insights. However, there are several factors touched on by the authors in the paragraph on their study's limitations that require further discussion.
First, being a case control study with no longitudinal data, there is limited interpretability of the effects of aging. The authors generalize, including all ages in their conclusion that normally the femoral neck expands in diameter faster in men than women. Our recent longitudinal study showed this not to be true for the seventh and eighth decades of age, when women show faster rates of subperiosteal and endosteal expansion, measured with the same software as Duan et al.(1, 2) We suggest that the weight of evidence links this reversal(3, 4) to postmenopausal estrogen status, because our population sample of women aged over 65 at recruitment were not taking hormone replacements, and our men likely had higher estrogen (as well as testosterone) levels than the women. As the authors mention, the evidence for an effect of estradiol at high pubertal levels in limiting sub-periosteal bone formation and increasing endosteal bone formation is strong. Lifelong physical activity, in contrast, is associated with increased femoral neck width.(5) Recent physical activity (not physical activity in the distant past) positively influences section modulus to increase resistance to bending of the hip.(5)
There are other concerns, mainly about the generalizability of the data on hip fracture. First, immobility per se, for example, after stroke, can be followed by rapid bone loss of up to 14% in the first year.(6) The level of recovery in mobility after hip fracture clearly might have influenced Duan et al's results in the individual case that was measured as long as 1 year after fracture. Compliance is an issue as well as are operator preferences concerning whom to scan; if the cases selected were above average in mobility after hip fracture (the usual effect of selection bias in hip fracture studies), it would be helpful to know. Second, there are some intriguing differences between Duan et al's results and our own case control study, which was done with different DXA equipment and a different hip strength software package in 68 female cases and 800 controls.(7) To enable comparison, we have calculated the section modulus Z-scores in the same way as Duan et al., from the subjects reported on by Crabtree et al.,(7) as well as examining the femoral neck width in cases and controls (we could not validly derive cortical thickness estimates using this software). The femoral neck diameter in our hip fracture cases was 0.4 mm less (not more) in these cases after adjusting for age, weight, and height (p = 0.02). Second, section modulus showed much greater discrimination in our study (mean Z-score for section modulus in fracture cases: −0.98; 95% CI, −1.24, −0.73). Despite the smaller number of cases in our study, the differences from Duan et al's results are striking.
Across Europe, we found that both BMD of the hip as well as geometrical determinants of hip strength varied substantially with geographical center,(8) suggesting that genetic or environmental determinants of the 10-fold variation in hip fracture rates across Europe(9) might in part be mediated through environmental or genetic effects on hip strength in old age. These considerations provoke several questions. Was case matching to controls, being different in our two studies, responsible for our different findings? Were exercise histories obtained in Duan et al's study? High levels in youth were not sustained through middle age in our experience(5) and could be a risk factor for a wide femoral neck mismatched to section modulus in old age. Was there any mismatch in the geographical origins of Duan et al's cases and controls? Many older Melbourne residents were born in old Europe. From a comparison of our two sets of data, it seems that the structural basis of femoral neck fragility may vary across cultures. In future, prospective studies of the evolution of hip strength, its determinants, and fracture outcomes may be needed before the contribution of femoral neck buckling to hip fracture risk can be quantitated in a fully generalizable way.